您当前的位置: > 详细浏览

变量相对重要性评估的方法选择及应用

Evaluation of predictors’ relative importance: Methods and applications

摘要:高维数据爆发的背景下,心理学研究目前急需变量相对重要性评估的有效方法。相对重要性评估的关键是选择合适的评估指标和统计推断方法。相对重要性的评估指标种类繁多,优势分析和相对权重是重点推荐的相对重要性评估指标。相对重要性的统计推断方法适用情境不同,Bootstrap抽样是推断单变量重要性和两变量重要性差异的常用方法,而贝叶斯检验是评估多变量重要性次序的新方法。线性回归模型之外,相对重要性研究已拓展到Logistic回归模型、结构方程模型、多水平模型等,但适用数据类型仍较为有限。相对重要性评估已广泛应用于心理学实证研究,但存在不恰当的指标解释和方法选择问题。为此,结合具体例子说明变量相对重要性的评估过程。

英文摘要:Evaluating predictors relative importance becomes increasingly important in the context of the explosion of high-dimensional data in psychological research. The key of relative importance analysis is to choose appropriate measures and inference approaches. Dominance analysis and relative weight are the recommended measures of relative importance among others. Bootstrap sampling is often used to infer the importance of a single variable or the difference between the importance of two variables. For three or more variables, Bayesian tests were recently developed to evaluate their importance orderings. Besides linear regression models, relative importance studies have been extended to logistic regression models, structural equation models, and multilevel models. However, only continuous predictors are concerned in these models. Although relative importance analysis has been widely used in psychological studies, researchers may incorrectly select and interpret the importance measures. Therefore, a real data example is used to illustrate how the relative importance can be evaluated.

版本历史

[V2] 2022-07-26 11:18:51 chinaXiv:202207.00039V2 下载全文
[V1] 2022-07-06 13:37:08 chinaXiv:202207.00039v1 查看此版本 下载全文
点击下载全文
同行评议状态
待评议
许可声明
metrics指标
  • 点击量2427
  • 下载量127
评论
分享
邀请专家评阅